价格计算方法论

1. 数据来源

1.1 交易所集成

CoinPaprika通过公开API与加密货币交易所集成。每家交易所必须提供市场交易对标识符（例如BTC/USD、ETH/BTC）、每个市场交易对的最新成交价格以及每个市场交易对的24小时交易量。

1.2 适用范围

价格计算管道中仅包含现货市场。衍生品市场（期货、永续合约、期权）因资金费率、到期机制和杠杆动态，其价格可能与标的现货价格大幅偏离，因此被排除在外。

1.3 外部参考数据

法定货币汇率来源于Open Exchange Rates，用于法定货币（EUR、GBP、JPY等）与USD之间的换算。稳定币锚定汇率（USDT、USDC及其他主要稳定币）使用从其USD入金市场衍生的成交量加权指数换算为USD。

2. 价格计算管道

该算法的设计完全基于当前市场状态计算价格，不依赖之前算法运行的价格。每次执行从零开始，从实时市场数据中自举自身的换算汇率，并通过多次迭代收敛到准确价格。整个管道每60秒执行一次。

1. 数据收集

   从每家集成交易所，CoinPaprika获取所有可用的现货市场数据。每个市场提供：市场交易对名称（例如BTC/USD）、计价货币的最新成交价格以及计价货币的24小时交易量。

2. 建立BTC/USD参考汇率

   算法从假定价格为零开始。第一个计算是BTC/USD加权平均价格，直接从所有BTC/USD现货市场推导得出。选择Bitcoin是因为BTC市场通常是各交易所中流动性最强的。

   BTC_USD = Σ(Price_i × Volume_i) / Σ(Volume_i)
   where i = each BTC/USD market across all exchanges

3. 第一轮：仅处理BTC和法定货币市场

   使用BTC/USD汇率，算法处理交易对一侧为BTC或法定货币的市场。以BTC计价的市场（例如ETH/BTC、SOL/BTC）：

   Price_USD = Price_BTC × BTC_USD

   以法定货币计价的市场（例如ETH/EUR、BTC/JPY）：

   Price_BTC = Price_USD / BTC_USD

4. 后续轮次：纳入所有市场

   许多加密货币也与其他加密货币进行交易。算法使用上一轮的换算汇率迭代地解析这些市场。

   Pass 1: BTC/USD markets → BTC_USD rate
   Pass 2: BTC + fiat markets → conversion rates for major coins
   Pass 3: All markets using Pass 2 rates → refined rates, more coins covered
   Pass N: Repeat until convergence → all reachable coins have prices

5. 异常值检测与剔除

   在最终汇总之前，识别并排除异常市场价格。预先排除的市场包括：零手续费交易、交易挖矿激励、过时数据（超过30分钟）及黑名单交易对。统计异常值检测使用根据流动性深度调整的基于中位数的边界。

   The statistical stages are described in Section 3.

6. 最终价格：成交量加权平均

   剔除异常值后，最终价格是所有剩余有效市场的成交量加权平均值：

   Price_USD = Σ(Price_i_USD × Volume_i_USD) / Σ(Volume_i_USD)

   流动性更高的市场自然获得更大权重，确保汇总价格反映大多数真实交易活动发生的位置。

3. Outlier Detection

Outlier removal runs as four independent stages before any price is averaged. Each catches a failure the others cannot see.

3.1 Stage 1 — Excluded Markets

Markets on exchanges under verification, and markets manually excluded after review, are dropped before anything is measured.

3.2 Stage 2 — Volume Anomalies

Volume is checked before price, because a fabricated volume corrupts every weight downstream. Markets reporting negative or implausibly large volume are rejected outright.

Past that bound, an asset's markets are accumulated smallest-first, and a market is rejected only when it fails several independent tests at once: the accumulated total must exceed a floor, the market's volume must be out of proportion to the smaller markets already accumulated, and the new total must exceed both the asset's previous observed maximum and its 24-hour median by a multiplier. Requiring all of them together is what keeps a genuine new listing on a large venue from being mistaken for fabricated volume. The limits scale with the asset's own median volume: the more liquid the asset, the tighter the constraint.

A rejection here is not final. Flagged markets are surfaced for review, and a market confirmed as genuine is re-admitted to the calculation.

3.3 Stage 3 — Trusted Price Band

Rather than applying one statistical test to every asset, CoinPaprika derives a trusted price band per asset per run, using the strongest method the available market depth supports and falling back through five tiers:

1. Interquartile. Requires several markets above a volume floor. Qualification is decided on the asset's stronger markets, but the band itself is computed across all but its emptiest ones. The band widens with the asset's own dispersion, so a volatile asset is not punished for being volatile.
2. Median bound, narrow. Requires several markets above a lower floor, with more than half of all markets already clustered around the median. The band is then tightened to the lowest and highest price actually observed.
3. Median bound, wide. As above, with fewer markets required and a wider tolerance.
4. Reference price. No usable market cohort, but a previous reference price exists. The band's width scales with the asset's median volume.
5. Last observed price. No reference price either, but a recent observation exists.

A liquid asset with many honest markets gets a tight band. A thin asset gets a wide band rather than a wrong one.

Stablecoins are treated as a separate class. The volume threshold for the strongest tier is lowered, so that tier stays available for assets whose individual markets are smaller, and the dispersion factor is tightened. An asset that should sit at its peg is given a far narrower tolerance than a volatile one.

Prices outside an asset's band never enter its average. Markets inside the band face one more check: a market — or a group of markets sitting on the same side of the median — whose distance-weighted volume would pull the average materially away from the median is excluded as well. And if every market for an asset lands outside its band, the markets are rejected rather than the band: the asset keeps its previous price instead of adopting a suspect one.

For thinly-listed assets, absolute sanity checks run alongside the band: a price or volume implausible on its face, or an implied market capitalization exceeding Bitcoin's, is rejected regardless of what the band would allow.

3.4 Stage 4 — Exchange-Level Anomalies

A final pass aggregates volume per exchange and compares it against that venue's own recent history — its previous maximum and its 24-hour median. The comparison is made only when the venue's market list is unchanged: a volume jump explained by newly added markets is not an anomaly, but the same jump on an unchanged market set is. This catches a venue-wide distortion that each individual market would survive.

3.5 Why a Band Is Derived for Every Asset

Assets are priced through one another, so one bad market propagates. An asset quoted at $1 on a single small market, joined by a second market of similar size quoting the inverse pair at $1,000, would carry that asset's reference price from $1 to roughly $500 on the next run — and from there into every asset quoted against it, and onward through the graph. Deriving a band for every asset, including those with too little data to measure precisely, breaks that chain at the first link.

4. 交易量计算

4.1 按交易所的交易量

每家交易所的24小时交易量是其所有活跃、非黑名单交易对经USD换算后的交易量总和。30分钟内未更新的交易对被排除。

Exchange_Volume = Σ(Volume_pair_i × Price_pair_i_USD)

4.2 按资产的交易量

单个资产的24小时交易量是所有集成交易所上该资产所有有效市场交易对的交易量总和。

4.3 全球交易量

全球加密货币交易量是所有按交易所交易量的总和。

5. 交易所置信度评分

每家交易所在0到100的量表上获得置信度评分，反映其报告数据的可靠性和质量。置信度评分影响价格计算管道中交易所数据的加权和过滤方式。

5.1 评估标准

5.2 置信度评分应用

• 处于审核中的交易所将完全从价格和交易量计算中排除。
• 置信度评分会定期审查和更新，也可能在重大事件（黑客攻击、监管行动、长期API中断等）发生时立即变更。
• 交易所数据通过市场交易对流动性指标（参见第6节）接受额外审查。

5.3 Estimated Real Volume（ERV）

Estimated Real Volume是2019年引入的用于应对虚假交易量和刷量交易的指标。ERV从特定交易所所有市场的流动性指标推导得出。每个市场报告的24小时交易量使用分配给其流动性指标类别（良好、平均或差）的权重进行调整。

6. 市场交易对流动性指标（圆点）

单个市场交易对基于订单簿价差（最优买价和卖价之间的百分比差异）、订单簿深度（中间价±1%和±10%范围内的价值）、交易量一致性及异常值检查，获得实时流动性评估。

6.1 指标类别

• 良好 — 流动性强，市场状况正常。
• 平均 — 市场状况受限或波动；流动性低于报告交易量所显示的水平。
• 差 — 相对于报告交易量流动性不足，或市场状况严重失衡。

未通过流动性检查的市场交易对在VWAP计算中被降低权重。如果没有其他来源可以计算某资产的价格和交易量，该交易对仍会被使用——并附有明确说明表示流动性无法得到验证。

7. 流通供应量

7.1 Proof-of-Work代币

对于PoW类加密货币（例如Bitcoin、Litecoin），流通供应量直接从区块浏览器API查询。数值经多个独立区块浏览器交叉验证，并随新区块挖出自动更新。

7.2 代币（基于智能合约）

对于ERC-20及类似代币，流通供应量计算方式如下：

Circulating Supply = Total Supply - Locked Tokens

Locked Tokens includes:
  - Team/founder vesting wallets
  - Treasury or reserve wallets
  - Burn addresses
  - Locked staking contracts

7.3 核验

流通供应量数据与项目团队披露的数据交叉核对，并在解锁事件、代币销毁或其他影响供应量的事件发生时更新。为防止被利用，特定阈值被有意保密。

8. 衍生指标

8.1 市值

Per-asset:  Market Cap = Current Price (USD) × Circulating Supply
Global:     Global Market Cap = Σ(Market Cap of all tracked assets)

8.2 历史最高价（ATH）

对于每种资产，CoinPaprika追踪有史以来记录的最高汇总价格、发生日期以及距ATH的当前距离：

ATH Distance (%) = (ATH Price − Current Price) / ATH Price × 100%

9. 上架标准

9.1 加密货币资产

要在CoinPaprika上上架，加密货币必须满足以下要求：

• 拥有清晰项目信息、功能正常的项目官网
• 拥有可验证链上数据的可用区块浏览器
• 在至少一家CoinPaprika集成交易所上活跃交易
• 透明的沟通渠道——项目维护公开频道（X/Twitter、Telegram、Discord），及时更新并回应社区问题

9.2 交易所

要作为数据源被集成，交易所必须：

• 运营具有可验证交易活动的功能性网站
• 提供符合CoinPaprika数据标准（市场交易对、价格、交易量）的公开 REST API
• 维护最新的API文档
• 为数据质量问题提供专门的联系方式

10. Asset Identifiers

Reference data is only usable inside a regulated workflow if it joins to the identifiers that workflow already runs on. CoinPaprika maps its asset IDs to external identifiers:

• DTI — Digital Token Identifier (ISO 24165), issued by the Digital Token Identifier Foundation. The ISO standard identifier for digital tokens, and the identifier regulatory reporting regimes reference for digital assets.
• ISIN — International Securities Identification Number (ISO 6166), which joins crypto positions to instrument records in existing back-office, custody and accounting systems.
• Provider IDs from CoinMarketCap, CoinGecko and CryptoCompare, for reconciliation and migration without re-mapping a universe by hand.

Mappings are available through the API ID Mappings endpoint: docs.coinpaprika.com — API ID Mappings

11. Independence and Conflicts of Interest

CoinPaprika is independently owned. It is not owned by, and holds no stake in, any cryptocurrency exchange, trading venue, market maker or token issuer. This distinguishes CoinPaprika from aggregators owned by the venues whose data they publish.

11.1 What a Commercial Relationship Can and Cannot Affect

CoinPaprika operates a paid expedited listing service (Fast Track) and carries advertising. Both are disclosed here, because a claim of independence is only useful if it is specific about its limits.

Fast Track buys moderator time: a submission and its metadata are reviewed and processed on a priority schedule instead of the standard queue. It buys nothing else.

No payment of any kind affects:

• the price calculated for an asset, which is produced by the pipeline described on this page and subject to the same outlier detection as every other listed asset;
• an exchange's Confidence Score;
• a market pair's Liquidity Indicator;
• ranking, which is derived from calculated market capitalization.

Listing is a decision about coverage. Pricing is a measurement. An asset admitted through Fast Track is measured by exactly the same method as any other: a thin asset receives a correspondingly wide trusted band and a low liquidity indicator.

Advertising is labeled as advertising and is never presented as data.

Questions about this methodology, including any concern about a potential conflict of interest, can be raised through our contact page: coinpaprika.com/contact

12. 更新频率